Discriminant analysis of clinical laboratory data for diagnosis and treatment of tuberculosis and lung cancer

The study is based on the clinical observation and examination of 33 patients with stages 1 or 2 lung cancer (Group 1), 53 with stage 3 lung cancer (Group 2), and 44 patients with pulmonary tuberculosis (Group 3). All the patients underwent surgical treatment. A control group comprised 50 apparently healthy individuals. The clinical laboratory studies included general peripheral blood analysis of ESR, the levels of hemoglobin, leukocytes, lymphocytes, total protein and albumin. ANOVA revealed a highly significant (p < 0.00004) differences in 5 of the 6 study laboratory parameters, variations in the level of total protein between the groups did not differ from that in the groups (p = 0.25). ANOVA revealed significant covariation of the levels of protein with ESR (p < 0.00004) and those of albumin (p < 0.00004). When variations in the latter were eliminated, the effect of the groups on total protein variation was highly significant (p = 0.0018). Discriminant analysis showed that the parameters studied were of diagnostic value in the differential diagnosis: probabilities of correct subdivision of patients ranged 72.6 to 78%, those of erroneous classification being 27.4 and 22%, respectively. The actual probabilities of misclassification were twofold lower. The problems in the employment of the discriminant procedure in the differential diagnosis of the study diseases using the specific contingent principle are discussed and a specific clinical observation is cited as an example.